Our current research demonstrates that searching the full-text of medical journals consistently yields a much higher percentage of relevant documents than searching their bibliographic surrogates. It also shows that full-text searches result in a significantly greater number of irrelevant items. A steadily increasing number of full-text biomedical databases, in both online and compact disk format, makes an investigation of better searching techniques for these databases an important area of research in medical information science. We propose to undertake such an investigation. By creating and analyzing careful records of file manipulations on two systems which support large databases of full- text biomedical journals, we shall develop searching techniques for eliminating irrelevant documents without significant loss of those which are relevant. We shall pay particular attention to the process of synonym aggregation, so important for retrieval of relevant documents in full-text files. By analyzing techniques successful on one system but not on the other in terms of system software and file structure, we shall delineate those design features most conducive to successful searching in the full-text databases.